A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.
Python artificial-intelligencedagdata-sciencedata-visualizationdataflowdeveloper-toolsmachine-learningnotebookspipelinepythonreactivesqlweb-app
Add Whisper sandbox example (#10994)
Trevor Manz728708aff963
diff --git a/examples/misc/pixi_whisper.py b/examples/misc/pixi_whisper.pynew file mode 100644index 0000000000..95805144b4--- /dev/null+++ b/examples/misc/pixi_whisper.py@@ -0,0 +1,97 @@+# /// script+# requires-python = ">=3.12,<3.13"+# dependencies = ["marimo>=0.25.0", "openai-whisper>=20250625"]+#+# [tool.pixi.workspace]+# channels = ["conda-forge"]+#+# [tool.pixi.dependencies]+# ffmpeg = "*"+#+# [tool.pixi.target.linux.dependencies]+# pytorch-gpu = "*"+#+# [tool.pixi.target.win.dependencies]+# pytorch-gpu = "*"+#+# [tool.pixi.target.osx.dependencies]+# pytorch-cpu = "*"+# ///++import marimo++__generated_with = "0.25.0"+app = marimo.App(width="compact")++[email protected](hide_code=True)+def _(mo):+ mo.md("""+ With [Pixi installed](https://pixi.prefix.dev/latest/installation/), run:++ ```bash+ marimo edit --sandbox=pixi examples/misc/pixi_whisper.py+ ```++ Choose a recording to transcribe with Whisper. Pixi installs CUDA-enabled+ PyTorch on Linux and Windows, and CPU PyTorch on macOS. The first run+ downloads the model weights.+ """)+ return+++def _():+ import torch+ import whisper++ device = "cuda" if torch.cuda.is_available() else "cpu"+ model = whisper.load_model("tiny.en", device=device)+ return model, whisper+++def _():+ import marimo as mo++ files = mo.ui.dropdown(+ {+ "The Ashiel Mystery": "https://librosa.org/data/audio/5703-47212-0000.ogg",+ "The Age of Chivalry": "https://librosa.org/data/audio/3436-172162-0000.ogg",+ "Sense and Sensibility": "https://librosa.org/data/audio/198-209-0000.ogg",+ },+ value="The Ashiel Mystery",+ label="Recording:",+ )+ files+ return files, mo+++def _(files, mo, whisper):+ audio = whisper.load_audio(files.value)+ mo.audio(audio, rate=16000)+ return (audio,)+++def _(audio, model):+ result = model.transcribe(audio, language="en", fp16=model.device.type == "cuda")+ result["text"]+ return++[email protected](hide_code=True)+def _(mo):+ mo.md("""+ Audio: [LibriSpeech](https://www.openslr.org/12/), distributed by+ [librosa](https://librosa.org/doc/main/recordings.html) under+ [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).+ *The Ashiel Mystery* is read by Garth Comira, *The Age of Chivalry*+ by Anders Lankford, and *Sense and Sensibility* by Heather Barnett.+ """)+ return+++if __name__ == "__main__":+ app.run()